Fig. 2

Comparison between microbiome variations during RR-1 to those during the STS-135 mission or induced by space-type radiation using STARMAPs. a Microbiome differences between RR-1 Flight and Ground samples were similar to microbiome differences between the flight and ground samples from the shuttle mission STS-135. A detailed description of STARMAPs is provided in the “Methods” section. Briefly, to compare group differences in the gut microbiome in one dataset to another, STARMAPs first performs PCA using samples from the first dataset alone (e.g., RR-1 data, left) and tests whether samples were segregated by the groups of interest (e.g., Flight vs. Ground) along the PCA axes. Then, samples from the second dataset (e.g., STS-135 data, RIGHT) are projected onto the same PCA axes and are tested for their group segregation along these PCA axes. As a third test, STARMAPs also evaluates the similarity in the directions of changes in the two datasets. It draws a line through the centers of the two groups of samples in each dataset to represent the group differences and tests whether the cosine of the angle, θ, between the lines in two datasets is significantly different from 0. When cosθ = 1, the microbiome changes in the two datasets are in exactly the same direction, and when cosθ = − 1, the microbiome changes in the two datasets are in exactly the opposite directions. Finally, STARMAPs uses an omnibus P value to summarize the three tests above to evaluate the overall similarity. Note that, while the graphs depict only PC1 and PC2, the tests were performed with all PCA axes. b Microbiome variations during RR-1 were compared to those induced by 10 days (left) or 30 days (right) of exposures to high-LET radiation exposure on Earth. c Microbiome variations in RR-1 mice were compared to those in rats exposed to low-LET radiation while fed on diets with either an adequate iron content (left) or a high-iron content (right). Note that STARMAPs uses random samplings from the Dirichlet distribution to estimate abundances of microbial taxa detected in one dataset but not the other. As a result, each time when comparing RR-1 data to another dataset, PCA of RR-1 samples gives very similar but not identical segregation patterns. Results shown here are at the species level, and similar results were observed at higher taxonomic levels as well (Additional file 1). Sample sizes of RR-1 data in a–c: Basal, n = 10; Vivarium, n = 8; Ground, n = 7; Flight, n = 6. Sample sizes of STS-135 data in a: Ground, n = 7; Flight, n = 6. Sample sizes of irradiated mice in b: n = 10 in each group. Sample sizes of irradiated rats in c: sham/Normal-Fe, n = 9; irradiated/Normal-Fe, n = 8; sham/High-Fe, n = 7; irradiated/High-Fe, n = 8